af_resample: preserve frame properties
[libav.git] / libavfilter / af_resample.c
1 /*
2 *
3 * This file is part of Libav.
4 *
5 * Libav is free software; you can redistribute it and/or
6 * modify it under the terms of the GNU Lesser General Public
7 * License as published by the Free Software Foundation; either
8 * version 2.1 of the License, or (at your option) any later version.
9 *
10 * Libav is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
13 * Lesser General Public License for more details.
14 *
15 * You should have received a copy of the GNU Lesser General Public
16 * License along with Libav; if not, write to the Free Software
17 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
18 */
19
20 /**
21 * @file
22 * sample format and channel layout conversion audio filter
23 */
24
25 #include "libavutil/avassert.h"
26 #include "libavutil/avstring.h"
27 #include "libavutil/common.h"
28 #include "libavutil/dict.h"
29 #include "libavutil/mathematics.h"
30 #include "libavutil/opt.h"
31
32 #include "libavresample/avresample.h"
33
34 #include "audio.h"
35 #include "avfilter.h"
36 #include "formats.h"
37 #include "internal.h"
38
39 typedef struct ResampleContext {
40 const AVClass *class;
41 AVAudioResampleContext *avr;
42 AVDictionary *options;
43
44 int64_t next_pts;
45
46 /* set by filter_frame() to signal an output frame to request_frame() */
47 int got_output;
48 } ResampleContext;
49
50 static av_cold int init(AVFilterContext *ctx, AVDictionary **opts)
51 {
52 ResampleContext *s = ctx->priv;
53 const AVClass *avr_class = avresample_get_class();
54 AVDictionaryEntry *e = NULL;
55
56 while ((e = av_dict_get(*opts, "", e, AV_DICT_IGNORE_SUFFIX))) {
57 if (av_opt_find(&avr_class, e->key, NULL, 0,
58 AV_OPT_SEARCH_FAKE_OBJ | AV_OPT_SEARCH_CHILDREN))
59 av_dict_set(&s->options, e->key, e->value, 0);
60 }
61
62 e = NULL;
63 while ((e = av_dict_get(s->options, "", e, AV_DICT_IGNORE_SUFFIX)))
64 av_dict_set(opts, e->key, NULL, 0);
65
66 /* do not allow the user to override basic format options */
67 av_dict_set(&s->options, "in_channel_layout", NULL, 0);
68 av_dict_set(&s->options, "out_channel_layout", NULL, 0);
69 av_dict_set(&s->options, "in_sample_fmt", NULL, 0);
70 av_dict_set(&s->options, "out_sample_fmt", NULL, 0);
71 av_dict_set(&s->options, "in_sample_rate", NULL, 0);
72 av_dict_set(&s->options, "out_sample_rate", NULL, 0);
73
74 return 0;
75 }
76
77 static av_cold void uninit(AVFilterContext *ctx)
78 {
79 ResampleContext *s = ctx->priv;
80
81 if (s->avr) {
82 avresample_close(s->avr);
83 avresample_free(&s->avr);
84 }
85 av_dict_free(&s->options);
86 }
87
88 static int query_formats(AVFilterContext *ctx)
89 {
90 AVFilterLink *inlink = ctx->inputs[0];
91 AVFilterLink *outlink = ctx->outputs[0];
92
93 AVFilterFormats *in_formats = ff_all_formats(AVMEDIA_TYPE_AUDIO);
94 AVFilterFormats *out_formats = ff_all_formats(AVMEDIA_TYPE_AUDIO);
95 AVFilterFormats *in_samplerates = ff_all_samplerates();
96 AVFilterFormats *out_samplerates = ff_all_samplerates();
97 AVFilterChannelLayouts *in_layouts = ff_all_channel_layouts();
98 AVFilterChannelLayouts *out_layouts = ff_all_channel_layouts();
99
100 ff_formats_ref(in_formats, &inlink->out_formats);
101 ff_formats_ref(out_formats, &outlink->in_formats);
102
103 ff_formats_ref(in_samplerates, &inlink->out_samplerates);
104 ff_formats_ref(out_samplerates, &outlink->in_samplerates);
105
106 ff_channel_layouts_ref(in_layouts, &inlink->out_channel_layouts);
107 ff_channel_layouts_ref(out_layouts, &outlink->in_channel_layouts);
108
109 return 0;
110 }
111
112 static int config_output(AVFilterLink *outlink)
113 {
114 AVFilterContext *ctx = outlink->src;
115 AVFilterLink *inlink = ctx->inputs[0];
116 ResampleContext *s = ctx->priv;
117 char buf1[64], buf2[64];
118 int ret;
119
120 if (s->avr) {
121 avresample_close(s->avr);
122 avresample_free(&s->avr);
123 }
124
125 if (inlink->channel_layout == outlink->channel_layout &&
126 inlink->sample_rate == outlink->sample_rate &&
127 (inlink->format == outlink->format ||
128 (av_get_channel_layout_nb_channels(inlink->channel_layout) == 1 &&
129 av_get_channel_layout_nb_channels(outlink->channel_layout) == 1 &&
130 av_get_planar_sample_fmt(inlink->format) ==
131 av_get_planar_sample_fmt(outlink->format))))
132 return 0;
133
134 if (!(s->avr = avresample_alloc_context()))
135 return AVERROR(ENOMEM);
136
137 if (s->options) {
138 AVDictionaryEntry *e = NULL;
139 while ((e = av_dict_get(s->options, "", e, AV_DICT_IGNORE_SUFFIX)))
140 av_log(ctx, AV_LOG_VERBOSE, "lavr option: %s=%s\n", e->key, e->value);
141
142 av_opt_set_dict(s->avr, &s->options);
143 }
144
145 av_opt_set_int(s->avr, "in_channel_layout", inlink ->channel_layout, 0);
146 av_opt_set_int(s->avr, "out_channel_layout", outlink->channel_layout, 0);
147 av_opt_set_int(s->avr, "in_sample_fmt", inlink ->format, 0);
148 av_opt_set_int(s->avr, "out_sample_fmt", outlink->format, 0);
149 av_opt_set_int(s->avr, "in_sample_rate", inlink ->sample_rate, 0);
150 av_opt_set_int(s->avr, "out_sample_rate", outlink->sample_rate, 0);
151
152 if ((ret = avresample_open(s->avr)) < 0)
153 return ret;
154
155 outlink->time_base = (AVRational){ 1, outlink->sample_rate };
156 s->next_pts = AV_NOPTS_VALUE;
157
158 av_get_channel_layout_string(buf1, sizeof(buf1),
159 -1, inlink ->channel_layout);
160 av_get_channel_layout_string(buf2, sizeof(buf2),
161 -1, outlink->channel_layout);
162 av_log(ctx, AV_LOG_VERBOSE,
163 "fmt:%s srate:%d cl:%s -> fmt:%s srate:%d cl:%s\n",
164 av_get_sample_fmt_name(inlink ->format), inlink ->sample_rate, buf1,
165 av_get_sample_fmt_name(outlink->format), outlink->sample_rate, buf2);
166
167 return 0;
168 }
169
170 static int request_frame(AVFilterLink *outlink)
171 {
172 AVFilterContext *ctx = outlink->src;
173 ResampleContext *s = ctx->priv;
174 int ret = 0;
175
176 s->got_output = 0;
177 while (ret >= 0 && !s->got_output)
178 ret = ff_request_frame(ctx->inputs[0]);
179
180 /* flush the lavr delay buffer */
181 if (ret == AVERROR_EOF && s->avr) {
182 AVFrame *frame;
183 int nb_samples = av_rescale_rnd(avresample_get_delay(s->avr),
184 outlink->sample_rate,
185 ctx->inputs[0]->sample_rate,
186 AV_ROUND_UP);
187
188 if (!nb_samples)
189 return ret;
190
191 frame = ff_get_audio_buffer(outlink, nb_samples);
192 if (!frame)
193 return AVERROR(ENOMEM);
194
195 ret = avresample_convert(s->avr, frame->extended_data,
196 frame->linesize[0], nb_samples,
197 NULL, 0, 0);
198 if (ret <= 0) {
199 av_frame_free(&frame);
200 return (ret == 0) ? AVERROR_EOF : ret;
201 }
202
203 frame->pts = s->next_pts;
204 return ff_filter_frame(outlink, frame);
205 }
206 return ret;
207 }
208
209 static int filter_frame(AVFilterLink *inlink, AVFrame *in)
210 {
211 AVFilterContext *ctx = inlink->dst;
212 ResampleContext *s = ctx->priv;
213 AVFilterLink *outlink = ctx->outputs[0];
214 int ret;
215
216 if (s->avr) {
217 AVFrame *out;
218 int delay, nb_samples;
219
220 /* maximum possible samples lavr can output */
221 delay = avresample_get_delay(s->avr);
222 nb_samples = av_rescale_rnd(in->nb_samples + delay,
223 outlink->sample_rate, inlink->sample_rate,
224 AV_ROUND_UP);
225
226 out = ff_get_audio_buffer(outlink, nb_samples);
227 if (!out) {
228 ret = AVERROR(ENOMEM);
229 goto fail;
230 }
231
232 ret = avresample_convert(s->avr, out->extended_data, out->linesize[0],
233 nb_samples, in->extended_data, in->linesize[0],
234 in->nb_samples);
235 if (ret <= 0) {
236 av_frame_free(&out);
237 if (ret < 0)
238 goto fail;
239 }
240
241 av_assert0(!avresample_available(s->avr));
242
243 if (s->next_pts == AV_NOPTS_VALUE) {
244 if (in->pts == AV_NOPTS_VALUE) {
245 av_log(ctx, AV_LOG_WARNING, "First timestamp is missing, "
246 "assuming 0.\n");
247 s->next_pts = 0;
248 } else
249 s->next_pts = av_rescale_q(in->pts, inlink->time_base,
250 outlink->time_base);
251 }
252
253 if (ret > 0) {
254 out->nb_samples = ret;
255
256 ret = av_frame_copy_props(out, in);
257 if (ret < 0) {
258 av_frame_free(&out);
259 goto fail;
260 }
261
262 out->sample_rate = outlink->sample_rate;
263 if (in->pts != AV_NOPTS_VALUE) {
264 out->pts = av_rescale_q(in->pts, inlink->time_base,
265 outlink->time_base) -
266 av_rescale(delay, outlink->sample_rate,
267 inlink->sample_rate);
268 } else
269 out->pts = s->next_pts;
270
271 s->next_pts = out->pts + out->nb_samples;
272
273 ret = ff_filter_frame(outlink, out);
274 s->got_output = 1;
275 }
276
277 fail:
278 av_frame_free(&in);
279 } else {
280 in->format = outlink->format;
281 ret = ff_filter_frame(outlink, in);
282 s->got_output = 1;
283 }
284
285 return ret;
286 }
287
288 static const AVClass *resample_child_class_next(const AVClass *prev)
289 {
290 return prev ? NULL : avresample_get_class();
291 }
292
293 static void *resample_child_next(void *obj, void *prev)
294 {
295 ResampleContext *s = obj;
296 return prev ? NULL : s->avr;
297 }
298
299 static const AVClass resample_class = {
300 .class_name = "resample",
301 .item_name = av_default_item_name,
302 .version = LIBAVUTIL_VERSION_INT,
303 .child_class_next = resample_child_class_next,
304 .child_next = resample_child_next,
305 };
306
307 static const AVFilterPad avfilter_af_resample_inputs[] = {
308 {
309 .name = "default",
310 .type = AVMEDIA_TYPE_AUDIO,
311 .filter_frame = filter_frame,
312 },
313 { NULL }
314 };
315
316 static const AVFilterPad avfilter_af_resample_outputs[] = {
317 {
318 .name = "default",
319 .type = AVMEDIA_TYPE_AUDIO,
320 .config_props = config_output,
321 .request_frame = request_frame
322 },
323 { NULL }
324 };
325
326 AVFilter ff_af_resample = {
327 .name = "resample",
328 .description = NULL_IF_CONFIG_SMALL("Audio resampling and conversion."),
329 .priv_size = sizeof(ResampleContext),
330 .priv_class = &resample_class,
331
332 .init_dict = init,
333 .uninit = uninit,
334 .query_formats = query_formats,
335
336 .inputs = avfilter_af_resample_inputs,
337 .outputs = avfilter_af_resample_outputs,
338 };